Transcriptomic analysis of cobalt stress in the marine yeast Debaryomyces hansenii

Abstract

The yeast Debaryomyces hansenii overproduces riboflavin upon exposure to subtoxic levels of cobalt (Co⁺²). However, mechanisms for survival have yet to be studied and have been hindered by D. hansenii's high genetic heterogeneity among strains. In this study, we used transcriptomic analyses and RNA-seq in order to identify differentially expressed genes in D. hansenii in response to cobalt exposure. Highly upregulated genes under this condition were identified to primarily comprise DNA damage and repair genes, oxidative stress response genes, and genes for cell wall integrity and growth. The main response of D. hansenii to heavy metal stress is the activation of non-enzymatic oxidative stress response mechanisms and control of biological production of reactive oxygen species. Our results indicate that D. hansenii does not seem to be pre-adapted to survive high concentrations of heavy metals. These organisms appear to possess genetic survival and detoxification mechanisms that enable the cells to recover from heavy metal stress.

Debaryomyces hansenii has been genetically assessed under exposure to toxic levels of the heavy metal cobalt in this investigation in order to better understand its survival mechanisms.

Graphical Abstract Figure.

Debaryomyces hansenii has been genetically assessed under exposure to toxic levels of the heavy metal cobalt in this investigation in order to better understand its survival mechanisms.

INTRODUCTION

Debaryomyces hansenii is an extremophilic yeast that has been studied for its osmotic stress tolerance, its lipid accumulation and storage, its ability to synthesize various chemicals and enzymes important for industrial applications, and its capability to survive toxic concentrations of the heavy metal cobalt (Breuer and Harms 2006; Seda-Miró et al. 2007; Arroyo Gonzalez et al. 2009). It is also a highly heterogeneous yeast strain, with two major varieties described, namely, var. hansenii and var. fabryi (Breuer and Harms 2006). Previous studies performed by our laboratory on 34 D. hansenii strains that assessed their phenotypic and genetic characterization when exposed to cobalt and saline stress revealed a large genetic divergence between strains (Seda-Miró unpublished data).

The aim of the present study was to elucidate the mechanisms of tolerance and recovery after cobalt exposure in D. hansenii strain J6. Heavy metals are considered trace elements needed for proper biological function of metabolic and signaling pathways (Valko, Morris and Cronin 2005). However, the function they have at trace levels becomes toxic at higher concentrations given their high redox activity and strong binding potential resulting in the inhibition of normal physiological processes (Nies 1999; Valko, Morris and Cronin 2005). Studies of the effect of oxidative stress in D. hansenii have shown that this yeast has a higher sensitivity to oxidative conditions but lack the molecular depth required to understand the mechanisms that take place when the yeast is under oxidative stress (Navarrete et al. 2009). Previous work from this laboratory (Seda-Miró et al. 2007) showed that Co (II) exposure decreased the growth of the cells and resulted in an overproduction of riboflavin. Although the amount of riboflavin produced after exposure to cobalt was found to be directly related to the metal concentration in the media, its role in the acclimation to heavy metal is unclear.

The characterization of differentially expressed genes by means of microarrays and real-time quantitative PCR in D. hansenii is hindered by the high genetic heterogeneity between strains. The genomic sequences of only two strains are so far available – CBS 767 and MTCC 234 (Dujon et al. 2004; Kumar et al. 2012; Keeling et al. 2014). Earlier studies in our laboratory showed a lack of cross-reactivity in microarrays made from the genome of CBS 767 with mRNA of the J6 strain (Arroyo Gonzalez et al. 2009). In order to overcome this constraint, a transcriptome analysis was carried out which utilized RNA-seq data to elucidate the response of D. hansenii to heavy metal stress. The results provide valuable insights into the genes implicated in the recovery of this organism from heavy metal stress. This study reveals that D. hansenii has the capability of surviving heavy metals exposure by activating mechanisms that reduce non-ezymatically and avert the normal metabolic production of free radical species.

MATERIALS AND METHODS

Yeast strain and growth conditions

Debaryomyces hansenii strain J6 was used in this study. It was isolated from a Swedish estuary and was a gift from Dr L. Adler of the University of Gotteburg, Sweden. The yeast was grown in 250 mL of YPD (1% yeast extract, 2% peptone, 2%dextrose) medium and shaken in an orbital shaker (150 rpm) at 25°C until late log phase was reached (A₆₀₀ = 9.5–10). Approximately 2 ml samples were collected for the control (0 h), centrifuged (5 min at 6K rpm), washed two times with distilled water, flash-frozen and stored at −80°C. 5 mM Co (II) (CoSO_4•5H₂O) was added to the remaining culture and incubated under the same conditions. Samples were collected at 0.5, 1.5 and 3 h after exposure to cobalt, following the same procedure as before. Cells were counted at each time point utilizing a hemocytometer, and viability was calculated as the ratio of cell number in each treatment to the cell number of the control (0 h). Given that 5 mM Co (II) resulted in approximately 60% growth inhibition in all treatments, this concentration was chosen to perform transcriptomics analysis.

RNA extraction, library construction, RNA-seq and mapping

Total RNA was isolated with the RNeasy Mini Kit (Qiagen, CA, USA) with an on-column DNase treatment as per manufacturer's instructions. RNA concentration was determined by measuring absorbance at 260 nm (Genesis 10S UV-Vis, Thermo Scientific), purity was analyzed by the ratio of readings at 260 nm and 280 nm (A₂₆₀/A₂₈₀), and integrity was checked with an agarose gel electrophoresis. RNA samples were sent to the National Center for Genome Resources (NCGR) (Santa Fe, NM, USA), as part of the Marine Microbial Eukaryote Transcriptome Sequencing Project (MMETSP), where mRNA enrichment, cDNA library preparation, adapter addition, size selection (250-350 bp), PCR and RNA-seq (Illumina HiSeq 2000) were performed (Keeling et al. 2014). Pre-assembly quality and quality trimming was also performed at NCGR with a cut-off at Q15 (as described by Keeling et al. 2014 ). The resulting RNA-seq reads were mapped to D. hansenii 767 (GCA_000006445.2) using Burrows–Wheeler Alignment (v 0.5.9) (Li and Durbin 2009) allowing for three mismatches per 100 bp read. The mapped reads were filtered to remove reads mapping to more than two positions in the genome. The transcriptome may be accessed through the CAMERA Data Distribution Center (http://camera.crbs.ucsd.edu/mmetsp/) and in the Sequence Read Archive under BioProject PRJNA231566.

Expression and functional annotation

Using Artemis, the number of mapped reads of each gene was quantified and normalized into RPKM (reads per kb per million reads). To adjust for the variability in sequencing depth using the upper quantile normalization procedure (Bullard et al. 2010). Fold change in gene expression was calculated as log₂ ratios with respect to the control sample (0 h). Expression profiles (functional clusters) were obtained with Spotfire DecisionSite v. 9.1.1 by k-means clustering. Using only those genes that had a 4-fold or more change in expression at any given time point 6 clusters were identified. These genes were assigned IDs from UniProt protein database, Ensembl (EMBL-EBI and Wellcome Trust Sanger Institute) and the Candida Genome Database (Inglis et al. 2012). Gene ontology (GO) classification was obtained through the Gene Ontology Consortium (Ashburner et al. 2000). Functional clustering of those genes was done using DAVID to extract biological meaning from the large gene/protein list.

RNA extraction, primer design and qRT-PCR

Total RNA for qRT-PCR was extracted as mentioned above. Two genes with induced expression were identified (DEHA2F05742G (Histidinol-phosphate aminotransferase) and DEHA2D13772G (ATP phosphoribosyltransferase)) and primers were designed for quantitative real time PCR (qRTPCR) (Table 1). Actin (ACT1, Accession No. XM_45873) from D. hansenii CBS767 was used as the normalizing gene (Dujon et al. 2004). qRT-PCR was performed as a one-step reaction using Brilliant II SYBR® Green QRT-PCR Master Mix Kit (Agilent Technologies, Santa Clara, CA, USA) in the Mx3005P analyzer (Stratagene, La Jolla, CA, USA). The qRT-PCR mixture (25 μl total per well) consisted of 12.5 μl of SYBR® Green QRT-PCR master mix, 0.25 pmol μl⁻¹ of each primer for gene Histidinol-phosphate aminotransferase, 0.250 pmol μl⁻¹ forward primer, 0.50 pmol μl⁻¹ reverse primer for gene ATP phosphoribosyltransferase and 0.50 pmol μl⁻¹ of each primer for actin; 0.375 μl of the reference dye (ROX diluted 1:500), 1 μl of RT/RNase block enzyme mixture, 350 ng of total RNA and RNase-free water to adjust the final volume to 25 μl. The cycling protocol was performed as a three-step method, which was initiated at 50°C for 30 min (reverse transcription step), 95°C for 10 min, the 40 cycles of 95°C for 30 s, 50°C (for all three primers) for 1 min and 72°C for 30 s. Specificity of the primers was confirmed with a dissociation curve which initiated at 95°C for 1 min, 55°C for 30 s and then a ramp in temperature to 95°C with a data acquisition at a rate of 0.2°C/s). To calculate gene abundance, the ΔΔC_T method was applied.

Table 1.

Primer information for genes used in qRT-PCR as an RNAseq validation method.

Gene	Forward primer (5′ to 3′)	Reverse primer (5′ to 3′)	Product size (bp)	Tm (°C)
Histidinol-phosphate aminotransferase	CTTTCAGATAGATGTGGATG	GATTCACTAGATGTTTCAC	222	55.5 (Fwd) 53.2 (Rev)
ATP phosphoribosyltransferase	GATTTAGTTGAAAGTGGTGA	AATAGTTGCAGATCTTCTA	248	55.8 (Fwd) 53.3 (Rev)
Actin	GATTATGAAGTGTGATGTC	TTAGAAACACTTATGATGAAC	286	53.1 (Fwd)
				54 (Rev)

RESULTS

Growth

Induction of cells with 5 mM Co (II) at late log phase reduced cell density to 55, 67 and 66% after 0.5, 1.5 and 3 h of exposure, respectively (Supporting Information Fig. S3). Because of the growth inhibition observed, this cobalt concentration was chosen to obtain an accentuated genetic response to cobalt stress.

RNA-seq and differential gene expression

Paired-end sequencing of RNA (Q15) from D. hansenii when exposed to 5 mM Co (II) identified a total of 4721 coding regions (17 Mb raw data), of which 471 had a change at any one time point of 4-fold or higher as compared to the unexposed control (Supporting Information Fig. S3 and Supporting Information Table S2). Approximately 40% of these genes were functionally annotated using the Uniprot public database, while the other 60% appear to be of unknown function. Among the 40% functionally annotated genes, 7% showed high homology, 67% showed homology and 26% showed weak homology to genes from the organisms that will be mentioned subsequently. Mapping and functional annotation revealed that approximately 58% of the identified genes have homologs to Saccharomyces cerevisiae and 20% to Candida albicans; the other 22% was homologous to other species, among them Schizosaccharomyces pombe, Yarrowia lipolytica, Pichia stipites and Aspergillus oryzae. Figures 1 and S4 (Supporting Information) depict down- and upregulated oscillation between significantly expressed genes (at least 4-fold change), where 0.5 h after exposure to cobalt 54 genes were upregulated, while 57 were downregulated; 1.5 h after exposure 55 genes were upregulated and 69 were downregulated; and 3 h after exposure 121 genes where upregulated while 127 gene were downregulated.

Figure 1.

Temporal fold change of differentially expressed genes in J6 at different times after exposure to 5 mM Co (II).

Temporal trend and GO

Six temporal trends were identified, in which 95 of the genes had a monotonically decreasing trend, 89 showed a monotonically decreased expression with delayed response, 70 showed decrease followed by increase, 68 had an oscillating trend, 108 had a monotonically increasing response and 41 showed an increase followed by a decrease (clusters 1–6 in Fig. 2, Supporting Information Table S1).

Figure 2.

(a) Temporal trends of the 471 genes with a fold change (log base 2) of four or more. C1, C2, C3 and C4 correspond to 0, 0.5, 1.5 and 3 h, respectively. (b) Number of genes in each temporal cluster. Colors correspond to the trends seen in Fig. 2.

Figure 2.

 (Continued)

Genes that were upregulated and downregulated at each time point were functionally clustered into the following categories: binding, cell part, transferase activity, molecular function, lipid metabolism, ion binding, amino acid metabolism, catalytic activity, physiological response to stimulus, transporter activity, oxidation-reduction process, protein metabolic process, carbohydrate metabolism and biological process (Fig. 3, Supporting Information Table S2). Generally speaking, a noticeable trend is seen, where 30 min after the initial exposure to Co (II), cells had recovered by upregulating the expression of genes for DNA synthesis and repair, non-enzymatic oxidative stress response and upregulating metabolic pathways that lower the production of superoxide. This response was sustained at both 1.5 and 3 h. After 3 hours of exposure to Co (II) recovery was accompanied by the upregulation of genes needed for cell division and mitosis. Genes for protein biosynthesis and ribosomal processes have been observed to be downregulated under stress conditions (Marks et al. 2008), for this reason details of these processes will not be discussed.

Figure 3.

Temporal and functional gene expression of J6 when exposed to 5 mM Co (II). Percentage of up- and downregulated genes based on function.

Validation of RNA-seq data by qRT-PCR

The two genes used for qRT-PCR validation, Histidinol-phosphate transaminase and a gene for an ATP phosphoribosyl transferase, were chosen for having the highest overall fold change in the monotonically increasing cluster. Histidinol-phosphate transaminase catalyzes the formation of L-glutamate, an amino acid important in the GABA shunt pathway (Coleman et al. 2001; Bach et al. 2009) while ATP phosphoribosyltransferase has a role in histidine metabolism that has a role in non-enzymatic oxidative stress response (Pearce and Sherman 1999). Results indicated very similar trends of the chosen genes when compared to the RNA-seq data (Fig. 4).

Figure 4.

Comparison of qRT-PCR expression pattern with RNA-Seq fold change for (a) gene DEHA2F05742G and (b) DEHA2D13772G.

DISCUSSION

Response to Co (II) was characterized by the upregulation of genes for DNA repair and synthesis, oxidative stress response and cell wall repair. The activation of these genes can be correlated to a survival and recovery process in which there is an initial shock, where DNA damage is sustained, followed by a response to protect the cells from oxidative stress caused by cobalt, and finally cell recovery from shock, as seen by the upregulation of mechanisms for cellular division and growth.

DNA synthesis and repair

Co (II) is known to cause genotoxic damage when cells are exposed to levels higher than those biologically required, as seen by the production of nucleobases in cells exposed to heavy metals such as cobalt (Valko, Morris and Cronin 2005). Heavy metals have the ability to produce reactive oxygen species, which oxidize DNA, lipids, proteins and carbohydrates (Valko, Morris and Cronin 2005; Michán et al. 2013). Overexpression of DNA repair genes (RAD52, DNA mismatch repair protein MSH2, SMC1, replication factor C subunit RFC5, MMS21, etc.) in response to cobalt exposure had a reduced expression with time, suggesting DNA damage repair (Supporting Information Table S1). On the other hand, PML2, a gene involved in DNA synthesis and repair, chromosome segregation, nuclear division and transcription regulation, showed a sustained increase in expression throughout the time points analyzed (Supporting Information Table S1). PLM2 functions in DNA synthesis and repair when induced either at START in the cell cycle or upon DNA damage (Gasch et al. 2001; Horak et al. 2002). Also, SYM1 that has been shown to be activated during stress conditions and is involved in maintaining structural and functional stability of the inner mitochondrial membrane, including mtDNA stability, was upregulated at all three time points (Dallabona et al. 2010). Dallabona et al. (2010) demonstrated that sym1Δ mutants show a defect in electron transport chain components. These components play an important role in the oxidative stress response of D. hansenii, as will be discussed subsequently.

Oxidative stress response

NADH production and electron leaks to oxygen as they are being transported from reduced substrates to complex IV in the electron transport chain, form reactive oxygen species during normal metabolic processes (Jamieson 1998; Tretter and Adam-vizi 2000; Lambert and Brand 2004; Sulahian et al. 2006). Under heavy metal exposure, mitochondrial membrane potential can be reduced by the lack of NADH produced within the Krebs cycle since aconitase, α-ketoglutarate dehydrogenase and succinate semialdehyde dehydrogenase (SSADH) are highly affected by oxidative stress (Tretter and Adam-vizi 2000). However, the loss of NADH can be compensated by the catabolism of γ-aminobutyric acid (GABA) and the GABA shunt pathway (Sulahian et al. 2006; Bach et al. 2009; Michaeli et al. 2011). The GABA shunt is a highly conserved pathway where decarboxylation of L-glutamate results in the production of GABA, which is then converted into succinate by means of glutamate decarboxylase, GABA transaminase and SSADH (Coleman et al. 2001; Bach et al. 2009). Genes for these enzymes are induced by UGA3, a GABA-dependent regulator (André 1990; Bach et al. 2009). The GABA shunt helps in recovering the loss in mitochondrial electron chain potential through the enzyme SSADH, which produces NADH or NADPH by oxidizing succinate semialdehyde into succinate (Coleman et al. 2001). Transcriptomic analysis of upregulated genes indicates that the TCA cycle in D. hansenii is highly affected by oxidative stress, but the affected steps are bypassed through the GABA shunt, as discussed below.

During exposure to Co (II), UGA3 was seen to be activated and overexpressed at all three time points studied, with a decrease in expression with time, but always over a 4-fold change (Supporting Information Table S1). Also, the general amino acid permease GAP1, which is GABA specific (André 1990; Bach et al. 2009), was shown to be upregulated at 1.5 and 3 h after exposure to Co (II) (Supporting Information Table S1). To further explain the GABA shunt in D. hansenii, Fig. 5, generated online via Interactive Pathways (Ipath) explorer v2 (Ogata et al. 1999; Kanehisa et al. 2014) shows the downregulated enzyme aconitase (aconitate hydratase, Table S1). As can be seen in Fig. 5, the TCA cycle cannot progress due to a lack of expression of an essential enzyme within the cycle. This enzyme does not significantly affect potential generation in the mitochondrial electron transport chain because NADH supply can be preserved by the GABA shunt (Tretter and Adam-vizi 2000).

Figure 5.

TCA cycle. Enzyme marked in red (cis-Aconitase) is shown to be downregulated in D. hansenii when exposed to 5 mM Co (II). Downregulation or inhibition of cis-Aconitase blocks the TCA cycle, thus providing evidence of D. hansenii's use of the GABA shunt (Ogata et al. 1999; Kanehisa et al. 2014).

Succinate dehydrogenase, another TCA cycle and electron transport chain (complex II) enzyme, was also found to be inhibited by oxidative stress. Superoxide production from the electron transport chain is achieved through electron leakage when these are being transported through the chain, localized to complex I and III (Lambert and Brand 2004), and by complex I and II NADH dehydrogenase activity (Lambert and Brand 2004). The downregulation of succinate dehydrogenase suggests that the cells have decreased superoxide production by normal metabolic processes. This reduction prevents the coupling of metabolic superoxide production with superoxide production caused by exposure to heavy metals, and the subsequent increase in oxidative stress.

In response to oxidative stress caused by heavy metal exposure, D. hansenii showed downregulation of enzymes required for the elimination of toxic radical species (Supporting Information Table S1). However, a non-enzymatic defense system is evidenced by the upregulation of glutatuione S-transferase 1, sulfiredoxin, lysine and histidine metabolism, and sphingosine N-acyltranseferase LAG1 (Supporting Information Table S1). Studies have shown that glutathione acts as a free radical scavenger (Jamieson 1998) and has been previously shown to be produced in D. hansenii under oxidative stress (Navarrete et al. 2009). Sulfiredoxin acts as an antioxidant that may be involved in repair and signaling pathways of protein damage caused by oxidation (Biteau, Labarre and Toledano 2003). Other important genes that were upregulated at the last two time points include genes for the biosynthesis of lysine and histidine which act in non-enzymatic oxidative stress response (Pearce and Sherman 1999). Within the mechanisms for non-enzymatic oxidative stress response, protein STB5, highly upregulated at the 3 h time point, is known to be upregulated as a response to oxidative stress and is a known regulator of genes related to the pentose phosphate pathway and genes for the production of NADPH (Larochelle et al. 2006). LAG1 that has a lipid metabolism function and is a ceramide synthesis component was also found to be upregulated in response to oxidative stress. Ceramides have been found to be induced under oxidative stress conditions (Van Brocklyn and Williams 2012).

Cell division and growth

Recovery from oxidative damage resulting from exposure to cobalt can be seen by the upregulation of genes for vegetative growth (Supporting Information Table S1). Specifically, upregulation of paxillin-like protein 1 (PXL1, a Rho-family GTPase required for vegetative growth and mating) was observed at all three time points. PXL1 is a cell growth mediator that is necessary for selection and maintenance of polarized growth sites in the actin cytoskeleton (Mackin, Sousou and Erdman 2004). Polarized cell growth is achieved through cascading events, where bud site selection occurs soon after localization of the signal (Mackin, Sousou and Erdman 2004). These steps can progress through BUD4, which controls bud placement in haploid cells and was also found to be upregulated (Mackin, Sousou and Erdman 2004). In this case, cell division seems to be achieved through vegetative growth and not by sporulation, as seen by the downregulation of SEF1 (Supporting Information Table S1).

Debaryomyces hansenii is a highly extremophilic yeast. However, unlike in the case of osmotolerance (Arroyo Gonzalez et al. 2009), resistance to concentrations of cobalt that are toxic to other similar organisms does not seem to be a pre-adapted characteristic. This is supported by the fact that there is an initial shock when D. hansenni is exposed to Co (II), as indicated by the observation that many of the genes that are upregulated several fold are involved in DNA synthesis and repair. Even though cells undergo shock, they respond rapidly to oxidative stress as a consequence of heavy metal exposure. Apart from activating mechanisms to eliminate free radical species non-enzymatically, D. hansenii also has the capability of averting natural metabolic processes that produce oxygen radicals. This response is so effective that an initial recovery of this organism is demonstrated by the upregulation of vegetative growth mechanisms as early as 3 h after exposure. It would be interesting to monitor growth after the 3 h time point in order to ascertain the time required for a complete recovery.

SUPPLEMENTARY DATA

Supplementary data are available at FEMSYR online.

FUNDING

This research was financially supported by a graduate fellowship to Gumá Cintrón with RISE 2BEST UPRM, grant number NIH-R25GM088023 from the National Institute of General Medical Sciences. RNA sequencing was performed by the National Center for Genome Resources (NCGR) in partnership with the Gordon and Betty Moore Foundation's Marine Microbiology Initiative (MMI) and Community Cyberinfrastructure for Advanced Microbial Ecology Research and Analysis (CAMERA) giving our laboratory the opportunity to participate in the Marine Microbial Eukaryote Transcriptome Sequencing Project (MMETSP). The computational resources were provided by the Minnesota Supercomputing Institute.

Conflict of interest. None declared.